On Wednesday, I attended the Enterprise Connect afternoon session on QoS & Network design for converged networks, hosted by John Bartlett of NetForecast. The speakers included representation from Cisco, Polycom, Apparent Networks, Netcraftsmen and our own Manfred Arndt from HP Networking’s Advanced Technology Group.
The session set out to demystify the things you need to support real-time multi-media traffic such as voice and video on your network, including QoS on LAN, WAN and the internet, SLAs, topology, mobility, and monitoring.
Virtualization and Cloud
The biggest difference I've noticed between VoiceCon last year and Enterprise Connect this year is the acceleration of the move from dedicated hardware to software applications running on industry standard servers and endpoints. This is apparent from the variety of sessions and keynotes covering UC in the context of both virtualization and cloud.
This isn't a big stretch from the past, if you think that most IP-PBXs are really just servers running a real-time operating system, with a media gateway providing the external interface to the PSTN. The key difference is that the line between UC&C and other business applications is starting to blur, due to the seamless integration of the user experience on the desktop, as well as a growing assortment of new endpoint devices. The same benefits derived from virtualizing your standard business applications also applies to UC&C. A single common pool of compute and storage hardware can be utilized across multiple applications allowing you to scale the service to the needs of your business, whilst simultaneously delivering real cost savings in infrastructure.
With virtualization, we are now seeing a trend to host these applications in your own data-center, purchase it as a cloud-based service, or choose a hybrid delivery model using HP's recently announced CloudSystem.
Dramatically Improving Collaboration
The convergence of voice along with other enterprise applications as software running in the data center offers the potential for dramatically improving user collaboration across both desktop and mobile devices. However, due to the peer-to-peer nature of real-time applications, there is a need to host a portion of the solution in the branch to insure remote survivability, minimize latency and optimize WAN usage by sending the media traffic via the most direct path. This is where the ability to host these types of applications within the network infrastructure, like HP AllianceONE partner applications, makes a lot of sense.
While the additional benefits offered by these rich collaboration tools is compelling, they can lead to significant increases in the demands on the network that some try to dismiss by simply throwing in extra bandwidth. Unfortunately it is never that simple - with oversubscription and mismatch in link-speeds able to cause buffer overruns, even on a lightly loaded network, due to the bursty nature of TCP traffic.
Manfred listed his best practices for QoS in the LAN as
- Provision sufficient bandwidth
- Control unnecessary traffic & don’t admit more real-time traffic than you provisioned (Call Admission Control)
- Prioritize time sensitive traffic, but don't make it more complex than it needs to be
- Monitor for changes due to new applications or shifts in traffic
I won't cover the remaining slides as they're fairly self explanatory, and available below - but if you have any questions, please post them below in the comments.
For more information